← ThoughtMetricCONTENT HISTORY

Update to ThoughtMetric

Snapshot Nov 21, 2026 · 06:12 UTC · version 1.0.0

Collection source: downloaded plugin package.

WHAT CHANGED

First saved snapshot

No earlier snapshot is available to establish a change.

Save ThoughtMetric and check future changes in Following.

Up to 3 plugins free · no card required.

Compare saved observations

Download comparison JSON
Full technical diff · 0 changed fields
Full snapshot data
{
  "description": "Answer questions about a ThoughtMetric store's attributed marketing performance — revenue, orders, ad spend, ROAS, CAC, new vs returning customers, products and individual orders by channel, campaign, ad set, ad or over time. Use when the user asks how their marketing, ads, channels or campaigns are performing.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 434
    }
  ],
  "name": "marketing-attribution",
  "skill_md_contents": "---\nname: marketing-attribution\ndescription: Answer questions about a ThoughtMetric store's attributed marketing performance — revenue, orders, ad spend, ROAS, CAC, new vs returning customers, products and individual orders by channel, campaign, ad set, ad or over time. Use when the user asks how their marketing, ads, channels or campaigns are performing.\n---\n\nUse this skill to answer questions about the user's marketing attribution data with the ThoughtMetric tools. Explicit instructions from the user take priority over this workflow.\n\n## Workflow\n\n1. **Pick the project.** Call `list_projects`. If the account has one project, use it. If it has several and the user didn't name one, ask which store they mean. Don't guess between projects. Reuse the chosen project for follow-up questions in the conversation.\n\n2. **Resolve the date range.** Convert the user's wording into `start_date` and `end_date` in `YYYY-MM-DD` format, using today's date for relative ranges:\n   - \"last month\" is the previous calendar month.\n   - \"last 30 days\" ends today and starts 29 days earlier.\n   - \"this quarter\" runs from the first day of the current quarter to today.\n   If the user gives no range, use the last 30 days and say so in your answer.\n\n3. **Check the metric and dimension combination.** Before using a metric and dimension combination for the first time, call `get_metric_dimension_map`.\n   - Every metric in one query must support every dimension used, so use the intersection.\n   - `time` can be added to any valid dimension or pair of dimensions.\n   - If a combination isn't supported, split it into separate queries rather than dropping part of the question.\n   - Use `list_available_metrics` or `list_available_dimensions` when you need a name you don't know.\n\n4. **Query aggregates with `query_metrics`.**\n   - Group by `channel`, `segment`, `campaign_name`, `adset_name`, `ad_name`, `product_name` and similar dimensions.\n   - For trends, group by `time` and set `interval` to `day`, `week`, `month` or `year`.\n   - To narrow to one ad platform or channel, filter on `channel` with its ID, for example `facebookad`, `googlead`, `tiktokad`, `email` or `organic_social`.\n   - For \"top\" or \"worst\" questions, use `sort`, `sort_dir` and `limit`.\n\n5. **Use `order_list` only for order-level questions.** These are questions like \"show me orders\", \"which orders came from email\" or \"what did this customer buy\". It returns individual orders with their channel journey.\n   - Keep `limit` to what the user asked for.\n   - Page with `offset` only when `has_more` is true and the user wants more.\n   - For totals and trends use `query_metrics` instead.\n\n## Reporting results\n\n- **Use only returned numbers.** Report only what the tools returned. Never invent, estimate or extrapolate figures. If a query returns no rows, say there's no data for that range or filter.\n- **Currency.** Revenue and spend are in the project's reporting currency, which the tools don't name. Don't add a currency symbol unless the user has told you the currency.\n- **Attribution model.** Attributed figures use the attribution model configured for the project in ThoughtMetric, so they match the user's dashboard. Don't describe the results as first-touch, last-touch or any other model unless the user tells you which one they use.\n- **Ratios.** Show ROAS as a multiple, for example 3.2x, and rates as percentages.\n- **Format.** For comparisons across several channels, campaigns or dates, present a short table, then one or two sentences on what stands out.\n- **Orders.** When listing orders, show each order's ID and customer email together. Merchants recognise orders by who placed them, and the email is the customer identifier their store uses. Don't use customer emails for anything beyond identifying the merchant's own orders to them.\n\n## Boundaries\n\n- **Read-only.** These tools can't pause or edit campaigns, change budgets, connect integrations or change settings. If asked, say ThoughtMetric is read-only here and suggest doing it in the ad platform or the ThoughtMetric dashboard.\n- **Connected projects only.** Only query projects returned by `list_projects`. If the user asks about a store that isn't in that list, explain that you can only see projects their ThoughtMetric account has access to.\n- **Errors.** If a tool returns an error about an invalid metric or dimension combination, check `get_metric_dimension_map` and retry with a supported combination. Don't retry the same query unchanged.\n"
}

SHA-256 of public snapshot: 6563a735080579b410935a74b486d4f2b796392110bbdfcfa9ac8332564d15d1